Hi all, for research purpose I am looking for some facts concerning inter-domain AS-path failures. Does anyone of you knows some cites or sources were I can get some numbers on that issue? I am aware that I might ask for network internals that are usually kept secret or at least not public ;-). Anyway if anyone could give me a hint where to find some information about : . Percentage of inter-domain AS-path failures (total amount of inter-domain failures) . How often do inter-domain AS-path failures occur? I would really appreciate! Btw. I do know the Sigcomm Paper "Understanding BGP Mis- configuration". cheers, - goetz -- Dipl. Inform Goetz Lichtwald University of Karlsruhe, Germany Institute of Telematics Zirkel 2, 76128 Karlsruhe Phone: +49 (0) 721 - 608 6404 Fax: +49 (0) 721 - 608 6789 PGP-Key: http://www.tm.uka.de/~lichtwald/mykey.asc
On Mon, Feb 28, 2005 at 03:09:31PM +0100, Goetz Lichtwald wrote:
for research purpose I am looking for some facts concerning inter-domain AS-path failures.
What are "inter-domain AS-path failures"? Please specify. Best regards, Daniel -- CLUE-RIPE -- Jabber: dr@cluenet.de -- dr@IRCnet -- PGP: 0xA85C8AA0
On Mon, 2005-02-28 at 15:28 +0100, Daniel Roesen wrote:
On Mon, Feb 28, 2005 at 03:09:31PM +0100, Goetz Lichtwald wrote:
for research purpose I am looking for some facts concerning inter-domain AS-path failures.
What are "inter-domain AS-path failures"? Please specify.
I regard an inter-domain as-path failure as a failure, where the BGP "hold-timer" expires (for whatever reason). kind regards, - goetz -- Dipl. Inform Goetz Lichtwald University of Karlsruhe, Germany Institute of Telematics Zirkel 2, 76128 Karlsruhe Phone: +49 (0) 721 - 608 6404 Fax: +49 (0) 721 - 608 6789 PGP-Key: http://www.tm.uka.de/~lichtwald/mykey.asc
On Mon, Feb 28, 2005 at 03:51:57PM +0100, Goetz Lichtwald wrote:
On Mon, 2005-02-28 at 15:28 +0100, Daniel Roesen wrote:
On Mon, Feb 28, 2005 at 03:09:31PM +0100, Goetz Lichtwald wrote:
for research purpose I am looking for some facts concerning inter-domain AS-path failures.
What are "inter-domain AS-path failures"? Please specify.
I regard an inter-domain as-path failure as a failure, where the BGP "hold-timer" expires (for whatever reason).
So you mean specifically BGP sessions timing out (session flaps)? Or do you actually mean any session shutdown (for whatever reason)? Best regards, Daniel -- CLUE-RIPE -- Jabber: dr@cluenet.de -- dr@IRCnet -- PGP: 0xA85C8AA0
On Mon, 2005-02-28 at 16:12 +0100, Daniel Roesen wrote:
On Mon, Feb 28, 2005 at 03:51:57PM +0100, Goetz Lichtwald wrote:
On Mon, 2005-02-28 at 15:28 +0100, Daniel Roesen wrote:
On Mon, Feb 28, 2005 at 03:09:31PM +0100, Goetz Lichtwald wrote:
for research purpose I am looking for some facts concerning inter-domain AS-path failures.
What are "inter-domain AS-path failures"? Please specify.
I regard an inter-domain as-path failure as a failure, where the BGP "hold-timer" expires (for whatever reason).
Sorry for being still too imprecise.
So you mean specifically BGP sessions timing out (session flaps)? Or do you actually mean any session shutdown (for whatever reason)?
I want know -- in real life -- how often BGP regards an AS-path to be broken, i.e. no KEEPALVIE messages can be exchanged -- at least that is what BGP "thinks". This should result in an "BGP session timing out" (your first point). From rom my understanding BGP tries to reestablish the BGP session for a certain time (ConnectRetry timer) and waits for an incoming BGP session request. So, concerning your second remark I am not quite sure if I got the point. Where is the difference between them? Is there a "meta-state" between a "session flap" and a "shut down"? Anyway, for me it is important to get a feeling of how many of those session flaps occur with a duration on less than appr. 10 minutes. kind regards, - goetz -- Dipl. Inform Goetz Lichtwald University of Karlsruhe, Germany Institute of Telematics Zirkel 2, 76128 Karlsruhe Phone: +49 (0) 721 - 608 6404 Fax: +49 (0) 721 - 608 6789 PGP-Key: http://www.tm.uka.de/~lichtwald/mykey.asc
Bon soir, Goetz Lichtwald wrote:
On Mon, Feb 28, 2005 at 03:09:31PM +0100, Goetz Lichtwald wrote:
for research purpose I am looking for some facts concerning inter-domain AS-path failures.
So you mean specifically BGP sessions timing out (session flaps)? Or do you actually mean any session shutdown (for whatever reason)?
I want know -- in real life -- how often BGP regards an AS-path to be broken, i.e. no KEEPALVIE messages can be exchanged -- at least that is what BGP "thinks". This should result in an "BGP session timing out" (your first point).
From rom my understanding BGP tries to reestablish the BGP session for a certain time (ConnectRetry timer) and waits for an incoming BGP session request. So, concerning your second remark I am not quite sure if I got the point. Where is the difference between them? Is there a "meta-state" between a "session flap" and a "shut down"?
There are two states where the BGP session is down: Active and Idle. Active means that the BGP speaker has started to connect the TCP socket to its neighbour, but the connection isn't established yet. Idle means that the BGP speaker is either sleeping for a while before it retries. It can also mean that the BGP speaker has decided that the TCP connection can't be formed at this time, due to there being no valid route to the neighbour for example. In addition to these two states, various vendors have added some administrative shutdown states, such as Idle (Admin) and Idle (PfxCnt). These are usually dead-end states that only a human operator can clear.
Anyway, for me it is important to get a feeling of how many of those session flaps occur with a duration on less than appr. 10 minutes.
In real life, BGP neighbour associations are TCP connections that have an average lifetime measured in weeks or months. -- Aleksi Suhonen
On Mon, 2005-02-28 at 21:03 +0200, Aleksi Suhonen wrote:
Anyway, for me it is important to get a feeling of how many of those session flaps occur with a duration on less than appr. 10 minutes.
In real life, BGP neighbour associations are TCP connections that have an average lifetime measured in weeks or months.
Does this mean that the BGP session is as stable as the TCP session concerning its lifetime? If so, I do not understand why "route Flap Damping" is widely used or am I wrong here? kind regards, - goetz -- Dipl. Inform Goetz Lichtwald University of Karlsruhe, Germany Institute of Telematics Zirkel 2, 76128 Karlsruhe Phone: +49 (0) 721 - 608 6404 Fax: +49 (0) 721 - 608 6789 PGP-Key: http://www.tm.uka.de/~lichtwald/mykey.asc
Does this mean that the BGP session is as stable as the TCP session concerning its lifetime? If so, I do not understand why "route Flap Damping" is widely used or am I wrong here?
do not confuse the lifetime of the control channel with that of the announcements/withdraws of the prefix-specific data payloads carried in that channel. randy
participants (4)
-
Aleksi Suhonen -
Daniel Roesen -
Goetz Lichtwald -
Randy Bush